Re: [Xpert]Re: [Dri-devel] r200 and libxaa

2002-10-20 Thread Mark Vojkovich
On Sun, 20 Oct 2002, Jens Owen wrote:

> Mark Vojkovich wrote:
> > On 21 Oct 2002, Michel Dänzer wrote:
> >>
> >>The problem appeared when I merged radeon driver fixes by Kevin E.
> >>Martin from XFree86 CVS to DRI CVS. They involved adding new fields to
> >>the XaaInfoRec struct; the fact that they were added in the middle of
> >>that struct caused the DRI binary snapshots, which were built against
> >>the new version of that struct but ran with an XAA module built against
> >>the old version, to crash and burn. I moved the new fields to the end of
> >>the struct, bumped the minor XAA version and added code to the radeon
> >>driver to only access the new fields if it's dealing with the new XAA.
> >>That worked as expected when I tested it here by copying libxaa.a from
> >>my Debian 4.2 installation into my DRI tree, which makes me suspect that
> >>the remaining problem with the snapshots is due to toolchain
> >>incompatibilities or something along those lines.
> >>
> > 
> > 
> >I don't know why people do this.  This change just broke NVIDIA's
> > and ATI's binary drivers too.  Any changes to structures that are
> > referenced by drivers need to go at the end of the structures.
> 
> Mark,
> 
> I'd like to see driver level binary compatability continue, too.  I'm 
> not certain if Michael's changes have made it back to XFree86's CVS.  I 
> posted to this thread (and copied Xpert) because I read some posts to 
> dri-devel that implied Michel's fixes weren't enough to restore driver 
> compatability (old drivers, new XFree86 core and modules).  If I'm 
> mistaken, please let us know.

   I just changed the header file in CVS to move the changes to the
end.  I think that's all that's needed to restore binary compatibility
with the old XAA module.   I don't see anything else in there that has
changed recently that would break things.


Mark.
___
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert



Re: [Xpert]Re: [Dri-devel] r200 and libxaa

2002-10-20 Thread Jens Owen
Mark Vojkovich wrote:

On 21 Oct 2002, Michel Dänzer wrote:



On Mon, 2002-10-21 at 02:05, Jens Owen wrote:


José Fonseca wrote:


On Wed, Oct 16, 2002 at 04:52:59PM +1000, Simon Bland wrote:



I've seen ppl mention the libxaa.a from Fonseca seems to fix the problem
with the screen blanking. Thing is I can't seem to find this libxaa..
Can anyone point me in the right direction for this? And can anyone
confirm that this does seem to fix the screen blanking problem with the
Radeon 8500.



Initially I hosted on my website because there was no certain that it
would help. But now that I know it does I've moved it to
http://dri.sf.net/snapshots/extras/libxaa.a.bz2 and wrote a README.Sig11
on http://dri.sf.net/snapshots/ .


Can anyone on Xpert comment on XAA binary incompatability?  Does this 
just affect Forward compatability, i.e. older X environments with newer 
binary drivers?

The problem appeared when I merged radeon driver fixes by Kevin E.
Martin from XFree86 CVS to DRI CVS. They involved adding new fields to
the XaaInfoRec struct; the fact that they were added in the middle of
that struct caused the DRI binary snapshots, which were built against
the new version of that struct but ran with an XAA module built against
the old version, to crash and burn. I moved the new fields to the end of
the struct, bumped the minor XAA version and added code to the radeon
driver to only access the new fields if it's dealing with the new XAA.
That worked as expected when I tested it here by copying libxaa.a from
my Debian 4.2 installation into my DRI tree, which makes me suspect that
the remaining problem with the snapshots is due to toolchain
incompatibilities or something along those lines.




   I don't know why people do this.  This change just broke NVIDIA's
and ATI's binary drivers too.  Any changes to structures that are
referenced by drivers need to go at the end of the structures.


Mark,

I'd like to see driver level binary compatability continue, too.  I'm 
not certain if Michael's changes have made it back to XFree86's CVS.  I 
posted to this thread (and copied Xpert) because I read some posts to 
dri-devel that implied Michel's fixes weren't enough to restore driver 
compatability (old drivers, new XFree86 core and modules).  If I'm 
mistaken, please let us know.

--
   /\
 Jens Owen/  \/\ _
  [EMAIL PROTECTED]  /\ \ \   Steamboat Springs, Colorado

___
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert


Re: [Xpert]Re: [Dri-devel] r200 and libxaa

2002-10-20 Thread Mark Vojkovich
On 21 Oct 2002, Michel Dänzer wrote:

> On Mon, 2002-10-21 at 02:05, Jens Owen wrote:
> > José Fonseca wrote:
> > > On Wed, Oct 16, 2002 at 04:52:59PM +1000, Simon Bland wrote:
> > > 
> > >> I've seen ppl mention the libxaa.a from Fonseca seems to fix the problem
> > >> with the screen blanking. Thing is I can't seem to find this libxaa..
> > >> Can anyone point me in the right direction for this? And can anyone
> > >> confirm that this does seem to fix the screen blanking problem with the
> > >> Radeon 8500.
> > >>
> > > 
> > > Initially I hosted on my website because there was no certain that it
> > > would help. But now that I know it does I've moved it to
> > > http://dri.sf.net/snapshots/extras/libxaa.a.bz2 and wrote a README.Sig11
> > > on http://dri.sf.net/snapshots/ .
> > 
> > Can anyone on Xpert comment on XAA binary incompatability?  Does this 
> > just affect Forward compatability, i.e. older X environments with newer 
> > binary drivers?
> 
> The problem appeared when I merged radeon driver fixes by Kevin E.
> Martin from XFree86 CVS to DRI CVS. They involved adding new fields to
> the XaaInfoRec struct; the fact that they were added in the middle of
> that struct caused the DRI binary snapshots, which were built against
> the new version of that struct but ran with an XAA module built against
> the old version, to crash and burn. I moved the new fields to the end of
> the struct, bumped the minor XAA version and added code to the radeon
> driver to only access the new fields if it's dealing with the new XAA.
> That worked as expected when I tested it here by copying libxaa.a from
> my Debian 4.2 installation into my DRI tree, which makes me suspect that
> the remaining problem with the snapshots is due to toolchain
> incompatibilities or something along those lines.
> 

   I don't know why people do this.  This change just broke NVIDIA's
and ATI's binary drivers too.  Any changes to structures that are
referenced by drivers need to go at the end of the structures.


Mark.
___
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert



Re: [Xpert]Re: [Dri-devel] r200 and libxaa

2002-10-20 Thread Mark Vojkovich
On Sun, 20 Oct 2002, Jens Owen wrote:

> José Fonseca wrote:
> > On Wed, Oct 16, 2002 at 04:52:59PM +1000, Simon Bland wrote:
> > 
> >> I've seen ppl mention the libxaa.a from Fonseca seems to fix the problem
> >> with the screen blanking. Thing is I can't seem to find this libxaa..
> >> Can anyone point me in the right direction for this? And can anyone
> >> confirm that this does seem to fix the screen blanking problem with the
> >> Radeon 8500.
> >>
> > 
> > Initially I hosted on my website because there was no certain that it
> > would help. But now that I know it does I've moved it to
> > http://dri.sf.net/snapshots/extras/libxaa.a.bz2 and wrote a README.Sig11
> > on http://dri.sf.net/snapshots/ .
> 
> Can anyone on Xpert comment on XAA binary incompatability?  Does this 
> just affect Forward compatability, i.e. older X environments with newer 
> binary drivers?

   I don't know why you'd be using a different XAA module in the first
place.  There haven't been any known bugs in XAA in a long time.

   I'm not sure we have any compatibility rules for anything other
than driver interfaces.  XAA interacts with more than the driver, it
interacts with other modules and the core server.  There may be some
explicit assumptions that you are not mixing core modules.  Precautions
have not been made to allow mixing of an XAA module with a server
of a different vintage.

   As for drivers, you can use a driver older than the server and
it will generally work.  You can use a driver newer than the server
only if the driver doesn't use any entry points or exports newer than
the server, and even if that part is safe you usually need to pass
the -ignoreABI flag.  But I don't think we have any goals of binary
compability between core modules.


Mark.


___
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert



Re: [Xpert]Re: [Dri-devel] r200 and libxaa

2002-10-20 Thread Michel Dänzer
On Mon, 2002-10-21 at 02:05, Jens Owen wrote:
> José Fonseca wrote:
> > On Wed, Oct 16, 2002 at 04:52:59PM +1000, Simon Bland wrote:
> > 
> >> I've seen ppl mention the libxaa.a from Fonseca seems to fix the problem
> >> with the screen blanking. Thing is I can't seem to find this libxaa..
> >> Can anyone point me in the right direction for this? And can anyone
> >> confirm that this does seem to fix the screen blanking problem with the
> >> Radeon 8500.
> >>
> > 
> > Initially I hosted on my website because there was no certain that it
> > would help. But now that I know it does I've moved it to
> > http://dri.sf.net/snapshots/extras/libxaa.a.bz2 and wrote a README.Sig11
> > on http://dri.sf.net/snapshots/ .
> 
> Can anyone on Xpert comment on XAA binary incompatability?  Does this 
> just affect Forward compatability, i.e. older X environments with newer 
> binary drivers?

The problem appeared when I merged radeon driver fixes by Kevin E.
Martin from XFree86 CVS to DRI CVS. They involved adding new fields to
the XaaInfoRec struct; the fact that they were added in the middle of
that struct caused the DRI binary snapshots, which were built against
the new version of that struct but ran with an XAA module built against
the old version, to crash and burn. I moved the new fields to the end of
the struct, bumped the minor XAA version and added code to the radeon
driver to only access the new fields if it's dealing with the new XAA.
That worked as expected when I tested it here by copying libxaa.a from
my Debian 4.2 installation into my DRI tree, which makes me suspect that
the remaining problem with the snapshots is due to toolchain
incompatibilities or something along those lines.

Hope this clarifies.


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast

___
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert